DEV Community

HopBase
HopBase

Posted on Edited on Originally published at hop-base.com

Claude Fable 5.1 API: 1M Context, $0.25/M Cache Reads and Migrating from Fable 5

The short answer: Claude Fable 5.1 (model ID claude-fable-5-1) is live on HopBase; check GET /v1/models for your key, and existing keys only need the model name changed. It is called through the native Anthropic Messages API at /v1/messages, so the Anthropic SDK and Claude Code work with zero code changes. It has a 1M context window and 128K max output, and adaptive thinking is always on: you steer reasoning depth with output_config.effort instead of a thinking parameter. The list price matches Fable 5 at $10 input and $50 output per million tokens, with cache writes at $12.50 (5 min) or $20 (1 hour), but cache reads drop to $0.25 from $1.00, so long sessions and agent loops that re-read the same prefix save the most. When migrating, note three changes: tool_choice any or tool returns 400, thinking blocks must be passed back unchanged with append-only history, and assistant prefill is not supported.

Anthropic's newest flagship, Claude Fable 5.1 (model ID claude-fable-5-1), is live on HopBase. Check GET /v1/models for your key — existing keys only need the model name changed.

Model specs

  • 1M context window, 128K max output: whole repositories, long documents, and long multi-turn conversations fit in a single request, with no need to split long outputs.
  • Adaptive thinking, always on: no thinking parameter needed; steer reasoning depth and spend with output_config.effort (low through max).
  • Built for long-horizon agent work: stronger than Fable 5 at long-running autonomous coding, document / spreadsheet / slide work, multi-step research, vision, and computer use. Same tokenizer as Fable 5, so token counts are essentially unchanged.
  • Native Anthropic Messages API: standard /v1/messages; the Anthropic SDK and Claude Code switch over with zero code changes.

Pricing: official list price

Item (per million tokens) Claude Fable 5.1 Claude Fable 5
Input $10 $10
Output $50 $50
Cache write (5 min / 1 hour) $12.50 / $20 $12.50 / $20
Cache read (hit) $0.25 (0.025×) $1.00

Cache reads cost a quarter of Fable 5's, so long sessions and agent loops that re-read the same prefix save the most. Billing is list price × actual usage, visible in the console in real time; for enterprise volume, request a quote from the pricing page.

Three things to check when migrating from Fable 5 / Opus 5

  • Forced tool use is gone: tool_choice any / tool returns 400. Use auto plus an instruction naming the tool, and strict: true or structured outputs when you need schema-valid JSON.
  • Thinking blocks are bound to the model: pass them back unchanged within a conversation and keep the history append-only — editing earlier turns invalidates earlier thinking blocks.
  • No assistant prefill: as with every Claude model since 4.6, control output format with structured outputs or the system prompt.

Three-minute quickstart

curl https://api.hop-base.com/v1/messages \
  -H "x-api-key: sk-your-key" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-fable-5-1",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Hello, introduce yourself"}]
  }'
Enter fullscreen mode Exit fullscreen mode

Full setup steps for Claude Code and the Anthropic SDK are in the Anthropic protocol guide. If you already have a key, just switch the model name to claude-fable-5-1.

Top comments (0)